[1999]-[BNP7787]-[Design]
Drug design and discovery can be a tedious task. It involves extensive data mining and evaluation of drug candidates. A good example of the use of supercomputing-based research is BioNumerik Pharmaceuticals.
The challenge in designing drugs is finding the best balance between drug strength and drug delivery, while reducing or eliminating the side effects on the human body. These side effects are commonly characterized by damage to the kidneys, bone marrow, nervous system and liver, as well as hair loss and vomiting. Long development cycles can not only make drugs very expensive for patients, they also delay a drug's availability.
BioNumerik simulates molecular interactions and drug transformations in the body using complex proprietary pharmaceutical software running on
parallel supercomputers.
Use of Parallelism
BioNumerik Pharmaceuticals uses Cray supercomputers with a drug development method that produces drugs at a fast rate. The computers simulate a drug's interactions with its biochemical target inside the body, and also all the systems that it encounters before reaching its target. This approach is called mechanism-based drug discovery.
Their method uses two CRAY T90™ series supercomputers, BioNumerik can process more than 26 billion calculations per second. For example, in the development of BNP7787, a new cancer-fighting BioNumerik drug currently in FDA trials, BioNumerik calculated more than 12 trillion possible chemical interactions.
Success
Drug development cycle, from discovery to testing, was reduced from +5 years to 18 months.
10 unique small-molecule drug classes have been developed for the treatment of cancer.
BioNumerik Pharmaceuticals CEO won Silicon Graphics/Cray Research Leadership Award for Breakthrough Science (June 3, 1997)
Weakness
The programming must encompass a shared and distributed memory system: two separate supercomputers, each with their own multiple processors. Since this design process is not ab initio, they have to rely on the quality of their proprietary databases and the freely accessible databases of chemical and biological data.
Challenges
The evaluation of 1 drug requires trillions of calculations.
Millions of possible drugs to screen.
Increasing screening accuracy increases computing time.
Pressure of market need for product.
Any reduction in overhead leads to dramatic reduction in product cost.
socrates.berkeley.edu |
